Network Working Group                                      A. Atlas, Ed.
Internet-Draft                                              Google, Inc.
Expires: August 5, 2006                                    A. Zinin, Ed.
                                                           February 2006

     Basic Specification for IP Fast-Reroute: Loop-free Alternates

Status of this Memo

   By submitting this Internet-Draft, each author represents that any
   applicable patent or other IPR claims of which he or she is aware
   have been or will be disclosed, and any of which he or she becomes
   aware will be disclosed, in accordance with Section 6 of BCP 79.

   Internet-Drafts are working documents of the Internet Engineering
   Task Force (IETF), its areas, and its working groups.  Note that
   other groups may also distribute working documents as Internet-

   Internet-Drafts are draft documents valid for a maximum of six months
   and may be updated, replaced, or obsoleted by other documents at any
   time.  It is inappropriate to use Internet-Drafts as reference
   material or to cite them other than as "work in progress."

   The list of current Internet-Drafts can be accessed at

   The list of Internet-Draft Shadow Directories can be accessed at

   This Internet-Draft will expire on August 5, 2006.

Copyright Notice

   Copyright (C) The Internet Society (2006).


   This document describes the use of loop-free alternates to provide
   local protection for unicast traffic in pure IP and MPLS/LDP networks
   in the event of a single failure, whether link, node or shared risk
   link group (SRLG).  The goal of this technology is to reduce the
   micro-looping and packet loss that happens while routers converge
   after a topology change due to a failure.  Rapid failure repair is
   achieved through use of precalculated backup next-hops that are loop-

Atlas, et al.            Expires August 5, 2006                 [Page 1]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

   free and safe to use until the distributed network convergence
   process completes.

Table of Contents

   1.  Introduction . . . . . . . . . . . . . . . . . . . . . . . . .  3
     1.1   Failure Scenarios  . . . . . . . . . . . . . . . . . . . .  5
   2.  Applicability of Described Mechanisms  . . . . . . . . . . . .  6
   3.  Alternate Next-Hop Calculation . . . . . . . . . . . . . . . .  7
     3.1   Basic Loop-free Condition  . . . . . . . . . . . . . . . .  8
     3.2   Node-Protecting Alternate Next-Hops  . . . . . . . . . . .  9
     3.3   Broadcast and NBMA Links . . . . . . . . . . . . . . . . .  9
     3.4   ECMP and Alternates  . . . . . . . . . . . . . . . . . . . 11
     3.5   Interactions with ISIS Overload, RFC 3137 and Costed
           Out Links  . . . . . . . . . . . . . . . . . . . . . . . . 11
       3.5.1   Interactions with ISIS Link Attributes . . . . . . . . 12
     3.6   Selection Procedure  . . . . . . . . . . . . . . . . . . . 12
   4.  Using an Alternate . . . . . . . . . . . . . . . . . . . . . . 16
     4.1   Terminating Use of Alternate . . . . . . . . . . . . . . . 17
   5.  Requirements on LDP Mode . . . . . . . . . . . . . . . . . . . 18
   6.  Routing Aspects  . . . . . . . . . . . . . . . . . . . . . . . 19
     6.1   Multi-Homed Prefixes . . . . . . . . . . . . . . . . . . . 19
     6.2   OSPF . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
       6.2.1   OSPF External Routing  . . . . . . . . . . . . . . . . 20
     6.3   BGP Next-Hop Synchronization . . . . . . . . . . . . . . . 21
     6.4   Multicast Considerations . . . . . . . . . . . . . . . . . 21
   7.  Security Considerations  . . . . . . . . . . . . . . . . . . . 21
   8.  IANA Considerations  . . . . . . . . . . . . . . . . . . . . . 21
   9.  References . . . . . . . . . . . . . . . . . . . . . . . . . . 22
     9.1   Normative References . . . . . . . . . . . . . . . . . . . 22
     9.2   Informative References . . . . . . . . . . . . . . . . . . 22
       Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . 23
   A.  OSPF Example Where LFA Based on Local Area Topology is
       Insufficient . . . . . . . . . . . . . . . . . . . . . . . . . 24
       Intellectual Property and Copyright Statements . . . . . . . . 26

Atlas, et al.            Expires August 5, 2006                 [Page 2]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

1.  Introduction

   Applications for interactive multimedia services such as VoIP and
   pseudo-wires can be very sensitive to traffic loss, such as occurs
   when a link or router in the network fails.  A router's convergence
   time is generally on the order of seconds; the application traffic
   may be sensitive to losses greater than 10s of milliseconds.

   As discussed in [I-D.ietf-rtgwg-ipfrr-framework], minimizing traffic
   loss requires a mechanism for the router adjacent to a failure to
   rapidly invoke a repair path, which is minimally affected by any
   subsequent re-convergence.  This specification describes such a
   mechanism which allows a router whose local link has failed to
   forward traffic to a pre-computed alternate until the router installs
   the new primary next-hops based upon the changed network topology.
   The terminology used in this specification is given in [I-D.ietf-
   rtgwg-ipfrr-framework].  The described mechanism assumes that routing
   in the network is performed using a link-state routing protocol--
   OSPF[RFC2328] or ISIS [RFC1195] [RFC2966].

   When a local link fails, a router currently must signal the event to
   its neighbors via the IGP, recompute new primary next-hops for all
   affected prefixes, and only then install those new primary next-hops
   into the forwarding plane.  Until the new primary next-hops are
   installed, traffic directed towards the affected prefixes is
   discarded.  This process can take seconds.

                        /------|  S  |--\
                       /       +-----+   \
                      / 5               8 \
                     /                     \
                   +-----+                +-----+
                   |  E  |                | N_1 |
                   +-----+                +-----+
                      \                     /
                  \    \  4              3 /  /
                   \|   \                 / |/
                   -+    \    +-----+    /  +-
                          \---|  D  |---/

                         Figure 1: Basic Topology

   The goal of IP Fast-Reroute is to reduce failure reaction time to 10s
   of milliseconds by using a pre-computed alternate next-hop, in the
   event that the currently selected primary next-hop fails, so that the

Atlas, et al.            Expires August 5, 2006                 [Page 3]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

   alternate can be rapidly used when the failure is detected.  A
   network with this feature experiences less traffic loss and less
   micro-looping of packets than a network without IPFRR.  There are
   cases where micro-looping is still a possibility since IPFRR coverage
   varies but in the worst possible situation a network with IPFRR is
   equivalent with respect to traffic convergence to a network without

   To clarify the behavior of IP Fast-Reroute, consider the simple
   topology in Figure 1.  When router S computes its shortest path to
   router D, router S determines to use the link to router E as its
   primary next-hop.  Without IP Fast-Reroute, that link is the only
   next-hop that router S computes to reach D. With IP Fast-Reroute, S
   also looks for an alternate next-hop to use.  In this example, S
   would determine that it could send traffic destined to D by using the
   link to router N_1 and therefore S would install the link to N_1 as
   its alternate next-hop.  At some later time, the link between router
   S and router E could fail.  When that link fails, S and E will be the
   first to detect it.  On detecting the failure, S will stop sending
   traffic destined for D towards E via the failed link, and instead
   send the traffic to S's pre-computed alternate next-hop, which is the
   link to N_1, until a new SPF is run and its results are installed.
   As with the primary next-hop, an alternate next-hop is computed for
   each destination.  The process of computing an alternate next-hop
   does not alter the primary next-hop computed via a standard SPF.

   If in the example of Figure 1, the link cost from N_1 to D increased
   to 30 from 3, then N_1 would not be a loop-free alternate, because
   the cost of the path from N_1 to D via S would be 17 while the cost
   from N_1 directly to D would be 30.  In real networks, we may often
   face this situation.  The existence of a suitable loop-free alternate
   next-hop is topology dependent and the nature of the failure the
   alternate is calculated for.

   A neighbor N can provide a loop-free alternate (LFA) if and only if

        Distance_opt(N, D) < Distance_opt(N, S) + Distance_opt(S, D)

                      Equation 1: Loop-Free Criterion

   A sub-set of loop-free alternate are downstream paths which must meet
   a more restrictive condition that is applicable to more complex
   failure scenarios:

                 Distance_opt(N, D) < Distance_opt(S, D)

                   Equation 2: Downstream Path Criterion

Atlas, et al.            Expires August 5, 2006                 [Page 4]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

1.1  Failure Scenarios

   The alternate next-hop can protect against a single link failure, a
   single node failure, one or more shared risk link group failure, or a
   combination of these.  Whenever a failure occurs that is more
   extensive than what the alternate was intended to protect, there is
   the possibility of temporarily looping traffic (note again, that such
   a loop would only last until the next complete SPF calculation).  The
   example where a node fails when the alternate provided only link
   protection is illustrated below.  If unexpected simultaneous failures
   occur, then micro-looping may occur since the alternates are not pre-
   computed to avoid the set of failed links.

   If only link protection is provided and the node fails, it is
   possible for traffic using the alternates to experience micro-
   looping.  This issue is illustrated in Figure 2.  If Link(S->E)
   fails, then the link-protecting alternate via N will work correctly.
   However, if router E fails, then both S and N will detect a failure
   and switch to their alternates.  In this example, that would cause S
   to redirect the traffic to N and N to redirect the traffic to S and
   thus causing a forwarding loop.  Such a scenario can arise because
   the key assumption, that all other routers in the network are
   forwarding based upon the shortest path, is violated because of a
   second simultaneous correlated failure - another link connected to
   the same primary neighbor.  If there are not other protection
   mechanisms a node failure is still a concern when only using link

                    +-----+       +-----+
                    |  S  |-------|  N  |
                    +-+---+   5   +-----+
                      |              |
                      | 5          4 |  |
                   |  |              | \|/
                  \|/ |              |
                      |    +-----+   |
                      +----|  E  |---+
                              | 10
                           |  D  |

Atlas, et al.            Expires August 5, 2006                 [Page 5]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

     Figure 2: Link-Protecting Alternates Causing Loop on Node Failure

   Micro-looping of traffic via the alternates caused when a more
   extensive failure than planned for can be prevented via selection of
   only downstream paths as alternates.  In Figure 2, S would be able to
   use N as an alternate, but N could not use S; therefore N would have
   no alternate and would discard the traffic, thus avoiding the micro-
   loop.  A micro-loop due to the use of alternates can be avoided by
   using downstream paths because each router in the path to the
   destination must be closer to the destination (according to the
   topology prior to the failures).  Although use of downstream paths
   ensures that the micro-looping via alternates does not occur, such a
   restriction can severely limit the coverage of alternates.

   It may be desirable to find an alternate that can protect against
   other correlated failures (of which node failure is a specific
   instance).  In the general case, these are handled by shared risk
   link groups (SRLGs) where any links in the network can belong to the
   SRLG.  General SRLGs may add unacceptably to the computational
   complexity of finding a loop-free alternate.

   However, a sub-category of SRLGs is of interest and can be applied
   only during the selection of an acceptable alternate.  This sub-
   category is to express correlated failures of links that are
   connected to the same router.  For example, if there are multiple
   logical sub-interfaces on the same physical interface, such as VLANs
   on an Ethernet interface, if multiple interfaces use the same
   physical port because of channelization, or if multiple interfaces
   share a correlated failure because they are on the same line-card.
   This sub-category of SRLGs will be referred to as local-SRLGs.  A
   local-SRLG has all of its member links with one end connected to the
   same router.  Thus, router S could select a loop-free alternate which
   does not use a link in the same local-SRLG as the primary next-hop.
   The local-SRLGs belonging to E can be protected against via node-
   protection; i.e. picking a loop-free node-protecting alternate.

   Where SRLG protection is provided, it is in the context of the
   particular OSPF or ISIS area, whose topology is used the SPF
   computations to compute the loop-free alternates.  If an SRLG
   contains links in multiple areas, then separate SRLG-protecting
   alternates would be required in each area that is traversed by the
   affected traffic.

2.  Applicability of Described Mechanisms

   IP Fast Reroute mechanisms described in this memo cover intra-domain
   routing only, with OSPF[RFC2328] or ISIS [RFC1195] [RFC2966] as the
   IGP.  Specifically, Fast Reroute for BGP inter-domain routing is not

Atlas, et al.            Expires August 5, 2006                 [Page 6]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

   part of this specification.

   Certain aspects of OSPF inter-area routing behavior explained in
   Section 6.2 and Appendix A impact the ability of the router
   calculating the backup next-hops to assess traffic trajectories.  In
   order to avoid micro-looping and ensure required coverage, certain
   constrains are applied to multi-area OSPF networks:

   a.  Loop-free alternates should not be used in the backbone area if
       there are any virtual links configured unless, for each transit
       area, there is a full mesh of virtual links between all ABRs in
       that area.  Loop-free alternates may be used in non-backbone
       areas regardless of whether there are virtual links configured.

   b.  Loop-free alternates should not be used for inter-area routes in
       an area that contains more than one alternate ABR.  [RFC3509]

   c.  Loop-free alternates should not be used for AS External routes or
       ASBR routes in a non-backbone area of a network where there
       exists an ABR that is announced as an ASBR in multiple non-
       backbone areas and there exists another ABR that is in at least
       two of the same non-backbone areas.

   d.  Loop-free alternates should not be used in a non-backbone area of
       a network for AS External routes where an AS External prefix is
       advertised with the same type of external metric by multiple
       ASBRs, which are in different non-backbone areas, with a
       forwarding address of or by one or more ASBRs with
       forwarding addresses in multiple non-backbone areas when an ABR
       exists simultaneously in two or more of those non-backbone areas.

3.  Alternate Next-Hop Calculation

   In addition to the set of primary next-hops obtained through a
   shortest path tree (SPT) computation that is part of standard link-
   state routing functionality, routers supporting IP Fast Reroute also
   calculate a set of backup next hops that are engaged when a local
   failure occurs.  These backup next hops are calculated to provide
   required type of protection (i.e. link-protecting and/or node-
   protecting) and to guarantee that when the expected failure occurs,
   forwarding traffic through them will not result in a loop.  Such next
   hops are called loop-free alternates or LFAs throughout this

   In general, to be able to calculate the set of LFAs for a specific
   destination D, a router needs to know the following basic pieces of

Atlas, et al.            Expires August 5, 2006                 [Page 7]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

   o  Shortest-path distance from the calculating router to the
      destination (Distance_opt(S, D))

   o  Shortest-path distance from the router's IGP neighbors to the
      destination (Distance_opt(N, D))

   o  Shortest path distance from the router's IGP neighbors to itself
      (Distance_opt(N, S))

   o  Distance_opt(S, D) is normally available from the regular SPF
      calculation performed by the link-state routing protocols.
      Distance_opt(N, D) and Distance_opt(N, S) can be obtained by
      performing additional SPF calculations from the perspective of
      each IGP neighbor (i.e. considering the neighbor's vertex as the
      root of the SPT--called SPT(N) hereafter--rather than the
      calculating router's one, called SPT(S)).

   This specification defines a form of SRLG protection limited to those
   SRLGs that include a link that the calculating router is directly
   connected to.  Information about local link SRLG membership is
   manually configured.  Information about remote link SRLG membership
   is dynamically obtained using [RFC4205] or [RFC4203].  In order to
   choose among all available LFAs that provide required SRLG protection
   for a given destination, the calculating router needs to track the
   set of SRLGs that the path through a specific IGP neighbor involves.
   To do so, each node D in the network topology is associated with
   SRLG_set(N, D), which is the set of SRLGs that would be crossed if
   traffic to D was forwarded through N. To calculate this set, the
   router initializes SRLG_set(N, N) for each of its IGP neighbors to be
   empty.  During the SPT(N) calculation, when a new vertex V is added
   to the SPT, its SRLG_set(N, V) is set to the union of SRLG sets
   associated with its parents, and the SRLG sets associated with the
   links from V's parents to V. The union of the set of SRLG associated
   with a candidate alternate next-hop and the SRLG_set(N, D) for the
   neighbor reached via that candidate next-hop is used to determine
   SRLG protection.

   The following sections provide information required for calculation
   of LFAs.  Sections Section 3.1 through Section 3.4 define different
   types of LFA conditions.  Section 3.5 describes constrains imposed by
   the IS-IS overload and OSPF stub router functionality.  Section 3.6
   defines the summarized algorithm for LFA calculation using the
   definitions in the previous sections.

3.1  Basic Loop-free Condition

   Alternate next hops used by implementations following this
   specification MUST conform to at least the loop-freeness condition

Atlas, et al.            Expires August 5, 2006                 [Page 8]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

   stated above in Equation 1.  This condition guarantees that
   forwarding traffic to an LFA will not result in a loop after a link

   Further conditions may be applied when determining link-protecting
   and/or node-protecting alternate next-hops as described in Sections
   Section 3.2 and Section 3.3.

3.2  Node-Protecting Alternate Next-Hops

   For an alternate next-hop N to protect against node failure of a
   primary neighbor E for destination D, N must be loop-free with
   respect to both E and D. In other words, N's path to D must not go
   through E. This is the case if Equation 3 is true, where N is the
   neighbor providing a loop-free alternate.

         Distance_opt(N, D) < Distance_opt(N, E) + Distance_opt(E, D)

      Equation 3: Criteria for a Node-Protecting Loop-Free Alternate

   If Distance_opt(N,D) = Distance_opt(N, E) + Distance_opt(E, D), it is
   possible that N has equal-cost paths and one of those could provide
   protection against E's node failure.  However, it is equally possible
   that one of N's paths goes through E, and the calculating router has
   no way to influence N's decision to use it.  Therefore, it SHOULD be
   assumed that an alternate next-hop does not offer node protection if
   Equation 3 is not met.

3.3  Broadcast and NBMA Links

   Verification of the link-protection property of a next hop in the
   case of a broadcast link is more elaborate than for a point-to-point
   link.  This is because a broadcast link is represented as a pseudo-
   node with zero-cost links connecting it to other nodes.

   Because failure of an interface attached to a broadcast segment may
   mean loss of connectivity of the whole segment, the condition
   described for broadcast link protection is pessimistic and requires
   that the alternate is loop-free with regard to the pseudo-node.
   Consider the example in Figure 3.

Atlas, et al.            Expires August 5, 2006                 [Page 9]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

                       +-----+    15
                       |  S  |--------
                       +-----+       |
                          | 5        |
                          |          |
                          | 0        |
                        /----\ 0 5 +-----+
                        | PN |-----|  N  |
                        \----/     +-----+
                           | 0        |
                           |          | 8
                           | 5        |
                        +-----+ 5  +-----+
                        |  E  |----|  D  |
                        +-----+    +-----+

           Figure 3: Loop-Free Alternate that is Link-Protecting

   In Figure 3, N offers a loop-free alternate which is link-protecting.
   If the primary next-hop uses a broadcast link, then an alternate
   SHOULD be loop-free with respect to that link's pseudo-node to
   provide link protection.  This requirement is described in Equation 4

              D_opt(N, D) < D_opt(N, pseudo) + D_opt(pseudo, D)

    Equation 4: Loop-Free Link-Protecting Criterion for Broadcast Links

   Because the shortest path from the pseudo-node goes through E, if a
   loop-free alternate from a neighbor N is node-protecting, the
   alternate will also be link-protecting unless the router S can only
   reach the alternate neighbor N via the same pseudo-node.  This can
   occur because S will direct traffic away from the shortest path to
   use an alternate.

   To obtain link protection, it is necessary both that the path from
   the selected alternate next-hop does not traverse the link of
   interest and that the link used from S to reach that alternate next-
   hop is not the link of interest.  The latter can only occur with non-
   point-to-point links.  Therefore, if the primary next-hop is across a
   broadcast or NBMA interface, it is necessary to consider link
   protection during the alternate selection.  Similar consideration of
   the link from S to the selected alternate next-hop as well as the
   path from the selected alternate next-hop is also necessary for SRLG

Atlas, et al.            Expires August 5, 2006                [Page 10]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

3.4  ECMP and Alternates

   With equal-cost multi-path, a prefix may have multiple primary next-
   hops that are used to forward traffic.  When a particular primary
   next-hop fails, alternate next-hops should be used to preserve the
   traffic.  These alternate next-hops may themselves also be primary
   next-hops, but need not be.  Other primary next-hops are not
   guaranteed to provide protection against the failure scenarios of

                           20 L1      L3  3
                        [N]-----[ S ]--------[E3]
                         |        |            |
                         |      5 | L2         |
                      20 |        |            |
                         |    ---------        | 2
                         |  5 |       | 5      |
                         |   [E1]    [E2]------|
                         |    |       |
                         | 10 |    10 |
                         |---[A]     [B]
                              |       |
                            2 |--[D]--| 2

     Figure 4: ECMP where Primary Next-Hops Provide Limited Protection

   In Figure 4 S has three primary next-hops to reach D; these are L2 to
   E1, L2 to E2 and L3 to E3.  The primary next-hop L2 to E1 can obtain
   link and node protection from L3 to E3, which is one of the other
   primary next-hops; L2 to E1 cannot obtain link protection from the
   other primary next-hop L2 to E2.  Similarly, the primary next-hop L2
   to E2 can only get node protection from L2 to E1 and can only get
   link protection from L3 to E3.  The third primary next-hop L3 to E3
   can obtain link and node protection from L2 to E1 and from L2 to E2.
   It is possible for both the primary next-hop L2 to E2 and the primary
   next-hop L2 to E1 to obtain an alternate next-hop that provides both
   link and node protection by using L1.

   Alternate next-hops are determined for each primary next-hop
   separately.  As with alternate selection in the non-ECMP case, these
   alternate next-hops should maximize the coverage of the failure

3.5  Interactions with ISIS Overload, RFC 3137 and Costed Out Links

   As described in [RFC3137], there are cases where it is desirable not
   to have a router used as a transit node.  For those cases, it is also
   desirable not to have the router used on an alternate path.

Atlas, et al.            Expires August 5, 2006                [Page 11]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

   For computing an alternate, a router MUST NOT use an alternate next-
   hop that is along a link whose cost or reverse cost is LSInfinity
   (for OSPF) or the maximum cost (for ISIS) or which has the overload
   bit set (for ISIS).  For a broadcast link, the reverse cost
   associated with a potential alternate next-hop is the cost towards
   the pseudo-node advertised by the next-hop router.  For point-to-
   point links, if a specific link from the next-hop router cannot be
   associated with a particular link, then the reverse cost considered
   is that of the minimum cost link from the next-hop router back to S.

   In the case of OSPF, if all links from router S to a neighbor N_i
   have a reverse cost of LSInfinity, then router S MUST NOT use N_i as
   an alternate.

   Similarly in the case of ISIS, if N_i has the overload bit set, then
   S MUST NOT consider using N_i as an alternate.

   This preserves the desired behavior of diverting traffic away from a
   router which is following [RFC3137] and it also preserves the desired
   behavior when an operator sets the cost of a link to LSInfinity for
   maintenance which is not permitting traffic across that link unless
   there is no other path.

   If a link or router which is costed out was the only possible
   alternate to protect traffic from a particular router S to a
   particular destination, then there should be no alternate provided
   for protection.

3.5.1  Interactions with ISIS Link Attributes

   [I-D.ietf-isis-link-attr] describes several flags whose interactions
   with LFAs needs to be defined.  A router SHOULD NOT specify the
   "local protection available" flag as a result of having LFAs.  A
   router SHOULD NOT use an alternate next-hop that is along a link for
   which the link has been advertised with the attribute "link excluded
   from local protection path" or with the attribute "local mainenance

3.6  Selection Procedure

   A router supporting this specification SHOULD attempt to select at
   least one loop-free alternate next-hop for each primary next-hop used
   for a given prefix.  A router MAY decide to not use an available
   loop-free alternate next-hop.  A reason for such a decision might be
   that the loop-free alternate next-hop does not provide protection for
   the failure scenario of interest.

   The alternate selection should maximize the coverage of the failure

Atlas, et al.            Expires August 5, 2006                [Page 12]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006


   When calculating alternate next-hops, the calculating router S
   applies the following rules.

   1.  S SHOULD select a loop-free node-protecting alternate next-hop,
       if one is available.  If no loop-free node-protecting alternate
       is available, then S MAY select a loop-free link-protecting

   2.  If S has a choice between a loop-free link-protecting node-
       protecting alternate and a loop-free node-protecting alternate
       which is not link-protecting, S SHOULD select a loop-free node-
       protecting alternate which is also link-protecting.  This can
       occur as explained in Section 3.3.

   3.  If S has multiple primary next-hops, then S SHOULD select as a
       loop-free alternate either one of the other primary next-hops or
       a loop-free node-protecting alternate if available.  If no loop-
       free node-protecting alternate is available and no other primary
       next-hop can provide link-protection, then S SHOULD select a
       loop-free link-protecting alternate.

   4.  Implementations SHOULD support a mode where other primary next-
       hops satisfying the basic loop-free condition and providing at
       least a minimal level of protection are preferred over any non-
       primary alternates.  This mode is provided to allow the
       administrator to preserve traffic patterns based on regular ECMP

   Following the above rules maximizes the level of protection and use
   of primary (ECMP) next-hops.

   Each next-hop is associated with a set of non-mutually-exclusive
   characteristics based on whether it is used as a primary next-hop to
   a particular destination D, and the type of protection it can provide
   relative to a specific primary next-hop E:

   Primary Path - The next-hop is used by S as primary.

   Loop-Free Node-Protecting Alternate - This next-hop satisfies
      Equation 1 and Equation 3.  The path avoids S, S's primary
      neighbor E, and the link from S to E.

   Loop-Free Link-Protecting Alternate -  This next-hop satisfies
      Equation 1 but not Equation 3.  If the primary next-hop uses a
      broadcast link, then this next-hop satisfies Equation 4.

Atlas, et al.            Expires August 5, 2006                [Page 13]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

   An alternate path may also provide none, some or complete SRLG
   protection as well as node and link or link protection.  For
   instance, a link may belong to two SRLGs G1 and G2.  The alternate
   path might avoid other links in G1 but not G2, in which case the
   alternate would only provide partial SRLG protection.

   Below is an algorithm that can be used to calculate loop-free
   alternate next-hops.  The algorithm is given for informational
   purposes and implementations are free to use any other algorithm as
   long as it satisfies the rules described above.

   The following procedure describes how to select an alternate next-
   hop.  The procedure is described to determine alternate next-hops to
   use to reach each router in the topology.  Prefixes that are
   advertised by a single router can use the alternate next-hop computed
   for the router to which they are attached.  The same procedure can be
   used to reach a prefix that is advertised by more than one router
   when the logical topological transformation described in Section 6.1
   is used.

   S is the computing router and has candidate next-hops H_1 through
   H_k.  N_i and N_j are used to refer to neighbors of S. For a next-hop
   to be a candidate, the next-hop must be associated with a bi-
   directional link, as is determined by the IGP.  For a particular
   destination router D, let S have already computed D_opt(S, D), and
   for each neighbor N_i, D_opt(N_i, D), D_opt(N_i, S), and D_opt(N_i,
   N_j), the distance from N_i to each other neighbor N_j, and the set
   of SRLGs traversed by the path D_opt(N_i, D).  S SHOULD follow the
   below procedure for every primary next-hop selected to reach D. This
   set of primary next-hops is represented P_1 to P_p.  This procedure
   finds the alternate next-hop(s) for P_i.

   First, initialize the alternate information for P_i as follows:

      P_i.alt_next_hops = {}
      P_i.alt_type = NONE P_i.alt_link-protect = FALSE
      P_i.alt_node-protect = FALSE
      P_i.alt_srlg-protect = {}

   For each candidate next-hop H_h,

   1.   Initialize variables as follows:

           cand_type = NONE
           cand_link-protect = FALSE
           cand_node-protect = FALSE
           cand_srlg-protect = {}

Atlas, et al.            Expires August 5, 2006                [Page 14]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

   2.   If H_h is P_i, skip it and continue to the next candidate next-

   3.   If is administratively allowed to be used as an

           and the cost of less than the maximum,
           and the reverse cost of H_h is less than the maximum,
           and H_h.neighbor is not overloaded (for ISIS),
           and is bi-directional,

        then H_h can be considered as an alternate.  Otherwise, skip it
        and continue to the next candidate next-hop.

   4.   If D_opt( H_h.neighbor, D) >= D_opt( H_h.neighbor, S) + D_opt(S,
        D), then H_h is not loop-free.  Skip it and continue to the next
        candidate next-hop.

   5.   cand_type = LOOP-FREE.

   6.   If H_h is a primary next-hop, set cand_type to PRIMARY.

   7.   If is not, set can_link-protect to TRUE.

   8.   If D_opt(H_h.neighbor, D) < D_opt(H_h.neighbor, P_i.neighbor) +
        D_opt(P_i.neighbor, D), set cand_node-protect to TRUE.

   9.   If the router considers SRLGs, then set the cand_srlg-protect to
        the set of SRLGs traversed on the path from S via P_i to
        P_i.neighbor.  Remove the set of SRLGs to which H_h belongs from
        cand_srlg-protect.  Remove from cand_srlg-protect the set of
        SRLGs traversed on the path from H_h.neighbor to D. Now
        cand_srlg-protect holds the set of SRLGs to which P_i belongs
        and that are not traversed on the path from S via H_h to D.

   10.  If cand_type is PRIMARY, the router prefers other primary next-
        hops for use as the alternate, and the P_i.alt_type is not
        PRIMARY, goto Step 18.

   11.  If cand_node-protect is TRUE and P_i.alt_node-protect is FALSE,
        goto Paragraph 18.

   12.  If cand_link-protect is TRUE and P_i.alt_link-protect is FALSE,
        goto Step 18.

   13.  If cand_srlg-protect has a better set of SRLGs than
        P_i.alt_srlg-protect, goto Step 18.

Atlas, et al.            Expires August 5, 2006                [Page 15]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

   14.  If cand_srlg-protect is different from P_i.alt_srlg-protect,
        then select between H_h and P_i.alt_next_hops based upon
        distance, IP addresses, or any router-local tie-breaker.  If H_h
        is preferred, then goto to Step 18.  Otherwise, skip H_h and
        continue to the next candidate next-hop.

   15.  Based upon the alternate types, the alternate distances, IP
        addresses, or other tie-breakers, decide if H_h is preferred to
        P_i.alt_next_hops.  If so, goto Step 18.

   16.  Decide if P_i.alt_next_hops is preferred to H_h.  If so, then
        skip H_h and continue to the next candidate next-hop.

   17.  Add H_h into P_i.alt_next_hops.  Set P_i.alt_type to the better
        type of H_h.alt_type and P_i.alt_type.  Continue to the next
        candidate next-hop.

   18.  Replace the P_i alternate next-hop set with H_h as follows:

           P_i.alt_next_hops = {H_h}
           P_i.alt_type = cand_type
           P_i.alt_link-protect = cand_link-protect
           P_i.alt_node-protect = cand_node-protect
           P_i.alt_srlg-protect = cand_srlg-protect

        Continue to the next candidate next-hop.

4.  Using an Alternate

   If an alternate next-hop is available, the router redirects traffic
   to the alternate next-hop in case of a primary next-hop failure as

   When a next-hop failure is detected via a local interface failure or
   other failure detection mechanisms (see [FRAMEWORK]), the router

   1.  Remove the primary next-hop associated with the failure.

   2.  Install the loop-free alternate calculated for the failed next-
       hop if it is not already installed (e.g. the alternate is also a
       primary next-hop).

   Note that the router MAY remove other next-hops if it believes (via
   SRLG analysis) that they may have been affected by the same failure,
   even if it is not visible at the time of failure detection.

Atlas, et al.            Expires August 5, 2006                [Page 16]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

   The alternate next-hop MUST be used only for traffic types which are
   routed according to the shortest path.  Multicast traffic is
   specifically out of scope for this specification.

4.1  Terminating Use of Alternate

   A router MUST limit the amount of time an alternate next-hop is used
   after the primary next-hop has become unavailable.  This ensures that
   the router will start using the new primary next-hops.  It ensures
   that all possible transient conditions are removed and the network
   converges according to the deployed routing protocol.

   A router that implements [I-D.ietf-rtgwg-microloop-analysis] SHOULD
   follow the rules given there for terminating the use of an alternate.

   It is desirable to avoid micro-forwarding loops involving S. An
   example illustrating the problem is given in Figure 5.  If the link
   from S to E fails, S will use N1 as an alternate and S will compute
   N2 as the new primary next-hop to reach D. If S starts using N2 as
   soon as S can compute and install its new primary, it is probable
   that N2 will not have yet installed its new primary next-hop.  This
   would cause traffic to loop and be dropped until N2 has installed the
   new topology.  This can be avoided by S delaying its installation and
   leaving traffic on the alternate next-hop.

                          |  N2 |--------   |
                          +-----+   1   |  \|/
                              |         |
                              |     +-----+    @@>  +-----+
                              |     |  S  |---------|  N1 |
                           10 |     +-----+   10    +-----+
                              |        |               |
                              |      1 |    |          |
                              |        |   \|/    10   |
                              |     +-----+            |  |
                              |     |  E  |            | \|/
                              |     +-----+            |
                              |        |               |
                              |      1 |  |            |
                              |        | \|/           |
                              |    +-----+             |
                              |----|  D  |--------------

      Figure 5: Example where Continued Use of Alternate is Desirable

   This is an example of a case where the new primary is not a loop-free

Atlas, et al.            Expires August 5, 2006                [Page 17]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

   alternate before the failure and therefore may have been forwarding
   traffic through S. This will occur when the path via a previously
   upstream node is shorter than the the path via a loop-free alternate
   neighbor.  In these cases, it is useful to give sufficient time to
   ensure that the new primary neighbor and other nodes on the new
   primary path have switched to the new route.

   If the newly selected primary was loop-free before the failure, then
   it is safe to switch to that new primary immediately; the new primary
   wasn't dependent on the failure and therefore its path will not have

   Given that there is an alternate providing appropriate protection and
   while the assumption of a single failure holds, it is safe to delay
   the installation of the new primaries; this will not create
   forwarding loops because the alternate's path to the destination is
   known to not go via S or the failed element and will therefore not be
   affected by the failure.

   An implementation SHOULD continue to use the alternate next-hops for
   packet forwarding even after the new routing information is available
   based on the new network topology.  The use of the alternate next-
   hops for packet forwarding SHOULD terminate:

   a.  if the new primary next-hop was loop-free prior to the topology
       change, or

   b.  if a configured hold-down, which represents a worst-case bound on
       the length of the network convergence transition, has expired, or

   c.  if notification of an unrelated topological change in the network
       is received.

5.  Requirements on LDP Mode

   Since LDP traffic will follow the path specified by the IGP, it is
   also possible for the LDP traffic to follow the loop-free alternates
   indicated by the IGP.  To do so, it is necessary for LDP to have the
   appropriate labels available for the alternate so that the
   appropriate out-segments can be installed in the forwarding plane
   before the failure occurs.

   This means that a Label Switched Router (LSR) running LDP must
   distribute its labels for the FECs it can provide to all its
   neighbors, regardless of whether or not they are upstream.
   Additionally, LDP must be acting in liberal label retention mode so
   that the labels which correspond to neighbors that aren't currently

Atlas, et al.            Expires August 5, 2006                [Page 18]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

   the primary neighbor are stored.  Similarly, LDP should be in
   downstream unsolicited mode, so that the labels for the FEC are
   distributed other than along the SPT.

   If these requirements are met, then LDP can use the loop-free
   alternates without requiring any targeted sessions or signaling
   extensions for this purpose.

6.  Routing Aspects

6.1  Multi-Homed Prefixes

   An SPF-like computation is run for each topology, which corresponds
   to a particular OSPF area or ISIS level.  The IGP needs to determine
   loop-free alternates to multi-homed routes.  Multi-homed routes occur
   for routes obtained from outside the routing domain by multiple
   routers, for subnets on links where the subnet is announced from
   multiple ends of the link, and for routes advertised by multiple
   routers to provide resiliency.

   Figure 6 demonstrates such a topology.  In this example, the shortest
   path to reach the prefix p is via E. The prefix p will have the link
   to E as its primary next-hop.  If the alternate next-hop for the
   prefix p is simply inherited from the router advertising it on the
   shortest path to p, then the prefix p's alternate next-hop would be
   the link to C. This would provide link protection, but not the node
   protection that is possible via A.

                      5   +---+  4   +---+  5  +---+
                    ------| S |------| A |-----| B |
                    |     +---+      +---+     +---+
                    |       |                    |
                    |     5 |                  5 |
                    |       |                    |
                  +---+ 5 +---+   5       7    +---+
                  | C |---| E |------ p -------| F |
                  +---+   +---+                +---+

                       Figure 6: Multi-homed prefix

   To determine the best protection possible, the prefix p can be
   treated in the SPF computations as a node with uni-directional links
   to it from those routers that have advertised the prefix.  Such a
   node need never have its links explored, as it has no out-going

   If there exist multiple multi-homed prefixes that share the same

Atlas, et al.            Expires August 5, 2006                [Page 19]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

   connectivity and the difference in metrics to those routers, then a
   single node can be used to represent the set.  For instance, if in
   Figure 6 there were another prefix X that was connected to E with a
   metric of 1 and to F with a metric of 3, then that prefix X could use
   the same alternate next-hop as was computed for prefix p.

   A router SHOULD compute the alternate next-hop for an IGP multi-homed
   prefix by considering alternate paths via all routers that have
   announced that prefix.

6.2  OSPF

   OSPF introduces certain complications because it is possible for the
   traffic path to exit an area and then re-enter that area.  This can
   occur whenever a router considers the same route from multiple areas.
   There are several cases where issues such as this can occur.  They
   happen when another area permits a shorter path to connect two ABRs
   than is available in the area where the LFA has been computed.  To
   clarify, an example topology is given in Appendix A.

   a.  Virtual Links: These allow paths to leave the backbone area and
       traverse the transit area.  The path provided via the transit
       area can exit via any ABR.  The path taken is not the shortest
       path determined by doing an SPF in the backbone area.

   b.  Alternate ABR[RFC3509]: When an ABR is not connected to the
       backbone, it considers the inter-area summaries from multiple
       areas.  The ABR A may determine to use area 2 but that path could
       traverse another alternate ABR B that determines to use area 1.
       This can lead to scenarios similar to that illustrated in
       Figure 7.

   c.  ASBR Summaries: An ASBR may itself be an ABR and can be announced
       into multiple areas.  This presents other ABRs with a decision as
       to which area to use.  This is the example illustrated in
       Figure 7.

   d.  AS External Prefixes: A prefix may be advertised by multiple
       ASBRs in different areas and/or with multiple forwarding
       addresses that are in different areas, which are connected via at
       least one common ABR.  This presents such ABRs with a decision as
       to which area to use to reach the prefix.

6.2.1  OSPF External Routing

   When a forwarding address is set in an OSPF AS-external LSA, all
   routers in the network calculate their next-hops for the external

Atlas, et al.            Expires August 5, 2006                [Page 20]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

   prefix by doing a lookup for the forwarding address in the routing
   table, rather than using the next-hops calculated for the ASBR.  In
   this case, the alternate next-hops SHOULD be computed by selecting
   among the alternate paths to the forwarding link(s) instead of among
   alternate paths to the ASBR.

6.3  BGP Next-Hop Synchronization

   Typically BGP prefixes are advertised with AS exit routers router-id,
   and AS exit routers are reached by means of IGP routes.  BGP resolves
   its advertised next-hop to the immediate next-hop by potential
   recursive lookups in the routing database.  IP Fast-Reroute computes
   the alternate next-hops to all IGP destinations, which include
   alternate next-hops to the AS exit router's router-id.  BGP simply
   inherits the alternate next-hop from IGP.  The BGP decision process
   is unaltered; BGP continues to use the IGP optimal distance to find
   the nearest exit router.  MBGP routes do not need to copy the
   alternate next hops.

   It is possible to provide ASBR protection if BGP selected selected a
   set of IGP next-hops and allowed the IGP to determine the primary and
   alternate next-hops as if the BGP route were a multi-homed prefix.
   This is for future study.

6.4  Multicast Considerations

   Multicast traffic is out of scope for this specification of IP Fast-
   Reroute.  The alternate next-hops SHOULD not be used for multi-cast
   RPF checks.

7.  Security Considerations

   The mechanism described in this document does not modify any routing
   protocol messages, and hence no new threats related to packet
   modifications or replay attacks are introduced.  Traffic to certain
   destinations can be temporarily routed via next-hop routers that
   would not be used with the same topology change if this mechanism
   wasn't employed.  However, these next-hop routers can be used anyway
   when a different topological change occurs, and hence this can't be
   viewed as a new security threat.

8.  IANA Considerations

   This document requires no IANA considerations.

9.  References

Atlas, et al.            Expires August 5, 2006                [Page 21]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

9.1  Normative References

   [RFC1195]  Callon, R., "Use of OSI IS-IS for routing in TCP/IP and
              dual environments", RFC 1195, December 1990.

   [RFC2328]  Moy, J., "OSPF Version 2", STD 54, RFC 2328, April 1998.

   [RFC3036]  Andersson, L., Doolan, P., Feldman, N., Fredette, A., and
              B. Thomas, "LDP Specification", RFC 3036, January 2001.

9.2  Informative References

              Vasseur, J. and S. Previdi, "Definition of an IS-IS Link
              Attribute sub-TLV", draft-ietf-isis-link-attr-01 (work in
              progress), May 2005.

              Shand, M. and S. Bryant, "IP Fast Reroute Framework",
              draft-ietf-rtgwg-ipfrr-framework-05 (work in progress),
              March 2006.

              Zinin, A., "Analysis and Minimization of Microloops in
              Link-state Routing Protocols",
              draft-ietf-rtgwg-microloop-analysis-01 (work in progress),
              October 2005.

   [RFC2966]  Li, T., Przygienda, T., and H. Smit, "Domain-wide Prefix
              Distribution with Two-Level IS-IS", RFC 2966,
              October 2000.

   [RFC3137]  Retana, A., Nguyen, L., White, R., Zinin, A., and D.
              McPherson, "OSPF Stub Router Advertisement", RFC 3137,
              June 2001.

   [RFC3209]  Awduche, D., Berger, L., Gan, D., Li, T., Srinivasan, V.,
              and G. Swallow, "RSVP-TE: Extensions to RSVP for LSP
              Tunnels", RFC 3209, December 2001.

   [RFC3509]  Zinin, A., Lindem, A., and D. Yeung, "Alternative
              Implementations of OSPF Area Border Routers", RFC 3509,
              April 2003.

   [RFC4203]  Kompella, K. and Y. Rekhter, "OSPF Extensions in Support
              of Generalized Multi-Protocol Label Switching (GMPLS)",
              RFC 4203, October 2005.

Atlas, et al.            Expires August 5, 2006                [Page 22]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

   [RFC4205]  Kompella, K. and Y. Rekhter, "Intermediate System to
              Intermediate System (IS-IS) Extensions in Support of
              Generalized Multi-Protocol Label Switching (GMPLS)",
              RFC 4205, October 2005.

Authors' Addresses

   Alia K. Atlas (editor)
   Google, Inc.
   1600 Amphitheatre Parkway
   Mountain View, CA  94043


   Alex Zinin (editor)
   701 E Middlefield Rd.
   Mountain View, CA  94043


   Raveendra Torvi
   Avici Systems, Inc.
   101 Billerica Avenue
   N. Billerica, MA  01862

   Phone: +1 978 964 2026

   Gagan Choudhury
   200 Laurel Avenue, Room D5-3C21
   Middletown, NJ  07748

   Phone: +1 732 420-3721

Atlas, et al.            Expires August 5, 2006                [Page 23]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

   Christian Martin
   iPath Technologies


   Brent Imhoff
   Juniper Networks
   1194 North Mathilda
   Sunnyvale, CA  94089

   Phone: +1 314 378 2571

   Don Fedyk
   Nortel Networks
   600 Technology Park
   Billerica, MA  01821

   Phone: +1 978 288 3041

Appendix A.  OSPF Example Where LFA Based on Local Area Topology is

   This appendix provides an example scenario where the local area
   topology does not suffice to determine that an LFA is available.  As
   described in Section 6.2, one problem scenario is for ASBR summaries
   where the ASBR is available in two areas via intra-area routes and
   there is at least one ABR or alternate ABR that is in both areas.
   The following Figure 7 illustrates this case.

Atlas, et al.            Expires August 5, 2006                [Page 24]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

                     [ F ]-----------[ C ]
                       |               |
                       |               | 5
                    20 |          5    |     1
                       |   [ N ]-----[ A ]*****[ F ]
                       |     |         #         *
                       |  40 |         # 50      *  2
                       |     |    5    #    2    *
                       |   [ S ]-----[ B ]*****[ G ]
                       |     |         *
                       |   5 |         * 15
                       |     |         *
                       |   [ E ]     [ H ]
                       |     |         *
                       |   5 |         * 10**
                       |     |         *
                       |---[ X ]-----[ASBR]

                       ----  Link in Area 1
                       ****  Link in Area 2
                       ####  Link in Backbone Area 0

      Figure 7: Topology with Multi-area ASBR Causing Area Transiting

   In Figure 7, the ASBR is also an ABR and is announced into both area
   1 and area 2.  A and B are both ABRs that are also connected to the
   backbone area.  S determines that N can provide a loop-free alternate
   to reach the ASBR.  N's path goes via A. A also sees an intra-area
   route to ASBR via Area 2; the cost of the path in area 2 is 30, which
   is less than 35, the cost of the path in area 1.  Therefore, A uses
   the path from area 2 and directs traffic to F. The path from F in
   area 2 goes to B. B is also an ABR and learns the ASBR from both
   areas 1 and area 2; B's path via area 1 is shorter (cost 20) than B's
   path via area 2 (cost 25).  Therefore, B uses the path from area 1
   that connects to S.

Atlas, et al.            Expires August 5, 2006                [Page 25]

Internet-Draft     draft-ietf-rtgwg-ipfrr-spec-base-05     February 2006

Intellectual Property Statement

   The IETF takes no position regarding the validity or scope of any
   Intellectual Property Rights or other rights that might be claimed to
   pertain to the implementation or use of the technology described in
   this document or the extent to which any license under such rights
   might or might not be available; nor does it represent that it has
   made any independent effort to identify any such rights.  Information
   on the procedures with respect to rights in RFC documents can be
   found in BCP 78 and BCP 79.

   Copies of IPR disclosures made to the IETF Secretariat and any
   assurances of licenses to be made available, or the result of an
   attempt made to obtain a general license or permission for the use of
   such proprietary rights by implementers or users of this
   specification can be obtained from the IETF on-line IPR repository at

   The IETF invites any interested party to bring to its attention any
   copyrights, patents or patent applications, or other proprietary
   rights that may cover technology that may be required to implement
   this standard.  Please address the information to the IETF at

Disclaimer of Validity

   This document and the information contained herein are provided on an

Copyright Statement

   Copyright (C) The Internet Society (2006).  This document is subject
   to the rights, licenses and restrictions contained in BCP 78, and
   except as set forth therein, the authors retain all their rights.


   Funding for the RFC Editor function is currently provided by the
   Internet Society.

Atlas, et al.            Expires August 5, 2006                [Page 26]